如何获取熊猫数据框中行值不为零的列数计数

Question

I have the following data:

device_id   class   Channel A   Channel B   Channel C   Channel D   Channel E   Channel F   Channel G   Channel H   Channel I   Channel J
28          S           2           4           23          45          6           6           8           9           0           0
54          P           34          56          21          0           76          45          0           0           0           0
97          S           24          45          76          0           0           35          76          87          6           20
22          V           0           0           32          76          89          0           0           0           0           0

The channels occur in groups as per a mapping which I have defined in a dictionary as below:

The dictionary:

di = {              
'S' : ['Channel A','Channel B'],                
'P' : ['Channel C','Channel D','Channel E'],                
'V' : ['Channel F','Channel G','Channel H','Channel I',' Channel J']
}

I need to count the number of channels being watched under each device row-wise from the pandas dataframe.

Expected output:

device_id   class   Channels_S  Channels_P  Channels_V
28           S          2           3           3
54           P          2           2           1
97           S          2           1           5
22           V          0           3           0

Can someone please guide me with this?

Answer 1

Here's a trick you can use :

mask = df.set_index(['device_id','class']) != 0

d1 = mask.groupby({i:k for k,v in di.items() for i in v},axis=1).sum()

ndf = d1.add_prefix('Channel_').reset_index()

   device_id class  Channel_P  Channel_S  Channel_V
0         28     S        3.0        2.0        3.0
1         54     P        2.0        2.0        1.0
2         97     S        1.0        2.0        5.0
3         22     V        3.0        0.0        0.0

Explanation :

1. mask will give you a dataframe with booleans with values not equal to zero true and the rest false. we are setting device and class as index so they will not be considered.

2. Expand the list inside the dict so we can group the columns based on the key {i:k for k,v in di.items() for i in v}

    { 'Channel_F': 'V', 'Channel_J': 'V', 'Channel_E': 'P', 
     'Channel_G': 'V', 'Channel_D': 'P', 'Channel_B': 'S', 
     'Channel_I': 'V', 'Channel_A': 'S', 'Channel_C': 'P', 
     'Channel_H': 'V' }

3. Groupby axis 1 and then sum.

4. Add prefix to columns and reset the index.